Spatially Resolved Stellar Kinematics from LEGA-C: Increased Rotational Support in z ~ 0.8 Quiescent Galaxies

SPATIALLY RESOLVED STELLAR KINEMATICS FROM LEGA-C: INCREASED ROTATIONAL SUPPORT IN Z∼0.8 QUIESCENT GALAXIES

Rachel Bezanson,^{1, 2} Arjen van der Wel,^{3, 4} Camilla Pacifici,⁵ Kai Noeske,⁴ Ivana Bariˇsi´c,⁴ Eric F. Bell,⁶ Gabriel B. Brammer,⁵Joao Calhau,⁷ Priscilla Chauke,⁴ Pieter van Dokkum,⁸ Marijn Franx,⁹ Anna Gallazzi,¹⁰

Josha van Houdt,⁴ Ivo Labb´e,⁹Michael V. Maseda,⁹ Juan Carlos Mu˜nos-Mateos,¹¹ Adam Muzzin,¹² Jesse van de Sande,¹³ David Sobral,^{7, 9} Caroline Straatman,³and Po-Feng Wu⁴

1Department of Physics and Astronomy and PITT PACC, University of Pittsburgh, Pittsburgh, PA, 15260, USA

2Department of Astrophysics, Princeton University, Princeton, NJ 08544, USA

3Sterrenkundig Observatorium, Universiteit Gent, Krijgslaan 281 S9, B-9000 Gent, Belgium

4Max-Planck Institut f¨ur Astronomie, K¨onigstuhl 17, D-69117, Heidelberg, Germany

5Space Telescope Science Institute, 3700 San Martin Drive, Baltimore, MD 21218, USA

6Department of Astronomy, University of Michigan, 1085 South University Ave., Ann Arbor, MI 48109, USA

7Department of Physics, Lancaster University, Lancaster LA1 4YB, UK

8Astronomy Department, Yale University, New Haven, CT 06511, USA

9Leiden Observatory, Leiden University, P.O.Box 9513, NL-2300 AA Leiden, The Netherlands

10INAF-Osservatorio Astrofisico di Arcetri, Largo Enrico Fermi 5, I-50125 Firenze, Italy

11European Southern Observatory, Alonso de Crdova 3107, Casilla 19001, Vitacura, Santiago, Chile

12Department of Physics and Astronomy, York University, 4700 Keele St., Toronto, Ontario, Canada, MJ3 1P3

13Sydney Institute for Astronomy, School of Physics, A28, The University of Sydney, NSW, 2006, Australia

(Accepted April 5, 2018)

Submitted to ApJ ABSTRACT

We present stellar rotation curves and velocity dispersion profiles for 104 quiescent galaxies at z = 0.6 − 1 from the Large Early Galaxy Astrophysics Census (LEGA-C) spectroscopic survey. Rotation is typically probed across 10-20kpc, or to an average of 2.7R_e. Combined with central stellar velocity dispersions (σ₀) this provides the first determination of the dynamical state of a sample selected by a lack of star formation activity at large lookback time.

The most massive galaxies (M?> 2 × 10¹¹M) generally show no or little rotation measured at 5kpc (|V5|/σ0 < 0.2 in 8 of 10 cases), while ∼64% of less massive galaxies show significant rotation. This is reminiscent of local fast- and slow-rotating ellipticals and implies that low- and high-redshift quiescent galaxies have qualitatively similar dynamical structures. We compare |V₅|/σ₀ distributions at z ∼ 0.8 and the present day by re-binning and smoothing the kinematic maps of 91 low-redshift quiescent galaxies from the CALIFA survey and find evidence for a decrease in rotational support since z ∼ 1. This result is especially strong when galaxies are compared at fixed velocity dispersion;

if velocity dispersion does not evolve for individual galaxies then the rotational velocity at 5kpc was an average of 94 ± 22% higher in z ∼ 0.8 quiescent galaxies than today. Considering that the number of quiescent galaxies grows with time and that new additions to the population descend from rotationally-supported star-forming galaxies, our results imply that quiescent galaxies must lose angular momentum between z ∼ 1 and the present, presumably through dissipationless merging, and/or that the mechanism that transforms star-forming galaxies also reduces their rotational support.

Keywords: galaxies:kinematics and dynamics - galaxies: high-redshift - galaxies: evolution

Corresponding author: Rachel Bezanson rachel.bezanson@pitt.edu

arXiv:1804.02402v1 [astro-ph.GA] 6 Apr 2018

1. INTRODUCTION

Massive galaxies exhibit significant angular momen- tum at z ∼ 2 (e.g., F¨orster Schreiber et al. 2009, 2011; Tacconi et al. 2013; van der Wel et al. 2014b;

van Dokkum et al. 2015; Wisnioski et al. 2015; Belli et al. 2017; Straatman et al. 2017). This statement is in contrast with massive galaxies in the local Uni- verse, which are predominantly elliptical, where even so-called “fast-rotators” exhibit significant dispersion support (e.g., Emsellem et al. 2007, 2011). This dis- crepancy necessitates an evolution in rotational support through cosmic time, however how and why this change occurred remains up for debate. One possibility is that the quenching process itself destroys organized rotation and/or the destruction of organized rotation is effec- tively what quenches galaxies (e.g.,Hopkins et al. 2008;

Martig et al. 2009). Alternatively the evolution could be more gradual, owing to subsequent minor or major merging (e.g., Naab et al. 2009; Hilz et al. 2013; Naab et al. 2014). We have evidence that the latter must play some role from the size evolution of quiescent, or

“non-star-forming”, galaxies, which grow significantly in size on average through cosmic time (e.g., Daddi et al.

2005;Toft et al. 2007;Trujillo et al. 2007;van Dokkum et al. 2008;van der Wel et al. 2008;Newman et al. 2012;

van der Wel et al. 2014a, and references within). This growth is most likely due to dissipationless minor merg- ing (e.g.,Bezanson et al. 2009;Hopkins et al. 2009;Naab et al. 2009; van Dokkum et al. 2010; Oser et al. 2011) which would diminish net angular momentum. For a given population of quiescent galaxies the rate of stel- lar rotation should then decrease with cosmic time, im- plying that high-redshift quiescent galaxies would show more rotation than their present-day counterparts.

A complicating factor is that the high- and low- redshift populations cannot be directly compared due to the increase in the number of quiescent galaxies with cosmic time, as galaxies cease to form stars, or “quench”.

This effect, often called progenitor bias, has been inves- tigated thoroughly as a potential driver of the empirical size evolution of quiescent galaxies as new and more ex- tended additions to the red sequence would drive evolu- tion in the average size-mass relations (e.g.,van der Wel et al. 2009;Valentinuzzi et al. 2010b,a; Poggianti et al.

2013; Carollo et al. 2013; Lilly & Carollo 2016; Fagioli et al. 2016;Williams et al. 2016). Although there is ev- idence for some evolution in the velocity dispersions of star-forming galaxies through cosmic time, these galax- ies have been shown to be primarily rotating disks since at least z ∼ 2 (e.g.,Wisnioski et al. 2015;Simons et al.

2017). Without any structural transformation, these new additions would also represent an influx of still- rotating quiescent galaxies as they have had less time since quenching to lose their angular momentum than their older counterparts. Therefore, the fraction of ro- tating galaxies in the quiescent population may increase

over cosmic time. One further level of complexity in this picture is that quenching of star formation may coincide with a change in dynamical structure, as suggested by the smaller relative sizes of post-starburst galaxies that constitute the newest additions to the high-redshift qui- escent population (Whitaker et al. 2012a; Yano et al.

2016).

So far, evolution in the shape distribution of quiescent galaxies has provided one of the strongest constraints on the evolution of angular momentum among the pop- ulation of quiescent galaxies. The emerging picture is that oblate, flat shapes are more common among high- redshift quiescent galaxies than in the present-day uni- verse (van der Wel et al. 2011; Chevance et al. 2012;

Chang et al. 2013). The (projected) shape of a galaxy is obviously only a crude proxy of dynamical structure, and even for large samples the necessary assumption was made that the population of high-redshift quiescent galaxies was composed of galaxies with the same intrin- sic shapes as today’s galaxies: oblate disks and triaxial spheroids. The relative numbers of both types were then inferred to change with redshift (Chang et al. 2013).

However, it is not self-evident that galaxy structures are the same at different cosmic epochs and the corre- spondence between global shape and kinematic proper- ties may well evolve. Therefore, it is essential to ob- tain spatially resolved kinematics of high-redshift quies- cent galaxies, which must be measured from stellar ab- sorption features. Currently, such direct evidence comes from small samples without uniform or necessarily rep- resentative selections. These include two examples of strongly lensed galaxies at z ∼ 2 (Newman et al. 2015;

Toft et al. 2017) and samples of 25 z ∼ 0.5 cluster (van der Marel & van Dokkum 2007;Moran et al. 2007) and z ∼ 1 field galaxies (van der Wel & van der Marel 2008).

The latter samples were selected on the basis of visual morphology, that is, a visual determination of the ab- sence of a disk-like structure, preventing a rigorous anal- ysis of the evolution of rotation among quiescent galaxies at different epochs. Finally,Belli et al.(2017) found in- direct evidence of evolution in the rotational support of quiescent galaxies from dynamical masses. In this pa- per we present a much larger sample of ∼100 galaxies at z ∼ 0.8 selected by their lack of star-formation ac- tivity and with high-quality stellar rotation curves from the Large Early Galaxy Astrophysics Census (LEGA-C) survey.

This paper begins in §2 with a brief description of the LEGA-C survey and the extraction of spatially re- solved stellar kinematics. In §3 we investigate the em- pirically derived rotational support of massive quies- cent galaxies at z ∼ 0.8 and the trends of that rota- tion with galaxy properties derived from imaging data.

In §4 we use stellar kinematics derived from the CAL- IFA DR3 dataset to assess the effects of seeing on the LEGA-C observations and study the redshift-evolution of the rotational support of massive quiescent galaxies.

Finally in §5, we conclude with a discussion of these re- sults in the context of models of galaxy evolution and other observational and theoretical studies. Throughout this paper we assume a standard concordance cosmology (H0= 70km s⁻¹Mpc⁻¹, ΩM = 0.3, and ΩΛ = 0.7).

2. LEGA-C DATA AND STELLAR KINEMATICS 2.1. The LEGA-C Spectroscopic Survey

The spectroscopic data included in this analysis are drawn from the first year data release of the Large Early Galaxy Astrophysics Census (LEGA-C) survey (van der Wel et al. 2016). This project is a 128-night ESO Public Spectroscopic survey of massive galaxies at 0.6 < z < 1.0 in the COSMOS field using VIMOS on the VLT. The LEGA-C Survey primary sample of

∼ 3000 galaxies is selected with a photometric or spec- troscopic redshift-dependent K-magnitude limit (K = 20.7−7.5×log((1+z)/1.8)), corresponding to a represen- tative sampling of galaxy colors down to log M?/M &

10.4. The defining, unique aspect of the LEGA-C spec- tra is the deep 20-hour long integration at a resolution of R = 2500 in the wavelength range ∼6300 − 8800˚A.

The first year dataset consists of 7 masks of roughly 130 galaxies in each mask with slits that are oriented in the N-S direction. The combined data yield the ex- tremely high signal-to-noise S/N ∼ 20˚A⁻¹in the contin- uum. The data reduction procedure is described byvan der Wel et al. (2016). Two-dimensional and extracted one-dimensional reduced spectra are publicly available via the ESO Science Archive Facility.

2.2. Photometry: Stellar Populations and Structures Additional ancillary data are available for the LEGA- C sample in the COSMOS field. Targeted galaxies are selected from the UltraVISTA version DR1 4.1 K- selected catalogs (Muzzin et al. 2013a). Rest-frame colors are calculated from the UltraVISTA photometry (McCracken et al. 2012;Muzzin et al. 2013a) using the EAZY (Brammer et al. 2008) code and fixing redshifts to the LEGA-C spectroscopic redshifts. Stellar population properties, most notably stellar masses, are determined from the UltraVISTA photometry using the FAST code (Kriek et al. 2009) and usingBruzual & Charlot (2003) stellar population libraries, adopting aChabrier (2003) Initial Mass Function (IMF),Calzetti et al.(2000) dust extinction, and exponentially declining star-formation rates. Although formal uncertainties on stellar masses are relatively low, systematics likely dominate and we adopt an uncertainty of 0.2 dex followingMuzzin et al.

(2009). Star-formation rates are estimated from the UV and IR (24µm from Spitzer-MIPS) luminosities, follow- ing Whitaker et al. (2012b). Morphological informa- tion is derived for all galaxies from COSMOS Hubble Space Telescope (HST) ACS F814W imaging (Koeke- moer et al. 2007; Massey et al. 2010), which is well matched to the rest-frame optical at this redshift. Best- fit S´ersic parameters, and uncertainties are derived for

all LEGA-C galaxies using GALFIT (Peng et al. 2002) and GALAPAGOS (Barden et al. 2012) following the procedures outlined in van der Wel et al. (2012) and van der Wel et al. (2016). The quoted measurement uncertainties of structural parameters do not include a number of systematic uncertainties and specifically do not account for covariance of parameters, which could dominate e.g. for S´ersic parameters. For visual presen- tation, we match the LEGA-C catalog to imaging from the first public data release of the Hyper-Suprime Cam Subaru Strategic Program (HSC-SSP), which includes deep grizy imaging in the COSMOS field (Aihara et al.

2017).

Stellar population and structural properties for the year one LEGA-C massive galaxies are shown in Figure 1. Symbol color indicates whether galaxies are catego- rized as quiescent (red) or star-forming (blue) based on their rest-frame U − V and V − J colors (upper left panel), adopting the Muzzin et al. (2013b) color cuts, which are specifically defined for the UltraVISTA pho- tometric catalogs used in determining rest-frame colors.

We note that although this selection does a very good job of identifying galaxies with quiescent stellar pop- ulations, there are a subset of galaxies in the current sample with clearly detected emission lines (see Figure 2for examples).

Horizontal lines indicate galaxies included in the full sample for which the semi-major axis is significantly in- clined with respect to the VIMOS slits (|P A| ≥ 45^o).

These objects are not considered in this paper, as the mismatch between the kinematic axis and the slit will prevent us from tracing stellar rotation in a straightfor- ward manner (e.g.,Weiner et al. 2006;Straatman et al.

2017). For this paper we focus on major axis kinemat- ics in the 104 quiescent galaxies for which the major axes are aligned to within |P A| < 45^o of the N-S slits (circles in Figure1). We note that the quoted position angles are photometric and the kinematic axes can also be misaligned with the photometry (e.g., Franx et al.

1989; Emsellem et al. 2007). Emsellem et al. (2007) demonstrated that for fast rotators this effect is mini- mal (. 10%), but kinematic and photometric position angles can be significantly misaligned, by up to ∼50% in the SAURON sample. HoweverKrajnovi´c et al.(2011) showed that for 90% of galaxies in the ATLAS3D sam- ple, the kinematic misalignment will be ≤ 15^o.

The top right panel of Figure1 shows the rest-frame U − V colors of the two populations as a function of stel- lar mass. The bottom left panel shows the effective ra- dius along the semi-major axis versus stellar mass, with the solid, red diagonal line indicating thevan der Wel et al. (2014a) size-mass relation for quiescent galaxies and dashed blue line for star-forming galaxies. Finally, the bottom right panel shows the specific star formation rate (sSFR) versus stellar mass. We note here that SFRs determined for quiescent galaxies are notably uncertain and as 24 micron flux may be undetected or ambiguous,

0.50 0.75 1.00 1.25 1.50 1.75 Rest-frame V-J

0.50 0.75 1.00 1.25 1.50 1.75 2.00 2.25 2.50

Rest-frame U-V

Star-forming Quiescent

|PA| 45^o

10.0 10.5 11.0 11.5

log Stellar Mass [M ] 0.50

0.75 1.00 1.25 1.50 1.75 2.00 2.25 2.50

Rest-frame U-V

10.0 10.5 11.0 11.5

log Stellar Mass [M ] 1.0

10.0

Effective Radius [kpc]

van der Wel+14 Relations Quiescent Star-Forming

10.0 10.5 11.0 11.5

log Stellar Mass [M ] 11

10 9 8

lo g sS FR [y r

]

Figure 1. Properties of the complete LEGA-C year one dataset. Symbol colors differentiate between star-forming (blue) and quiescent (red) galaxies as determined by U-V and V-J rest-frame colors and cuts fromMuzzin et al.(2013b) (upper left panel).

Misaligned galaxies (|P A| ≥ 45^o) are excluded from this study and are indicated by horizontal lines. Star-forming and quiescent galaxies in the LEGA-C sample have different distributions in color (upper right panel), physical size (bottom left panel), and sSFR (bottom right panel); for this study we focus on the kinematics of the quiescent population.

these sSFRs are likely to be upper limits for our sam- ple of galaxies. The quiescent and star-forming galaxies in the LEGA-C sample exhibit different distributions in all four phase spaces, although the populations overlap slightly in all but the U-V and V-J colors, which are used to initially differentiate between them.

2.3. Spatially Resolved Stellar Kinematics We measure the stellar and gas phase line-of-sight kinematics for each galaxy using the Penalized Pixel- Fitting (pPXF) method (Cappellari & Emsellem 2004)

with the updated Python routines (Cappellari 2017).

For each two-dimensional LEGA-C spectrum, each row with median S/N > 2 per pixel is fit with two tem- plate sets that are allowed to independently shift and broaden: stellar population templates to fit the contin- uum and a collection of possible emission lines to fit the ionized gas emission. The stellar template is a linear, op- timal non-negative combination ofVazdekis(1999) sin- gle stellar population (SSP) models, which are based on the Medium resolution INT Library of Empirical

-4" -2" 0" 2" 4"

-4"

-2"0"2"4"

10 kpc HST ACS F814W

-4" -2" 0" 2" 4"

-4"

-2"0"2"4" HSC gri

5 0 5

Arcseconds ^{H10 H9 H8 K H Hδ G Hγ Hβ}

ID: 117400 Mask: 4 at z=0.6687

0 20 40 60 80 100 120

Flux [10−19ergs−1cm2Å−1]

Full (continuum+emission) Model Continuum Model

6500 7000 7500 8000 8500

Wavelength [Å]

3020 10010 2030 40

Residuals

-4" -2" 0" 2" 4"

-4"

-2"0"2"4"

10 kpc HST ACS F814W

-4" -2" 0" 2" 4"

-4"

-2"0"2"4" HSC gri

5 0 5

Arcseconds

[OII] [NeIII]H8 H² Hγ

H10 H9 H8 K H Hδ G Hγ

ID: 126950 Mask: 2 at z=0.8840

0 10 20 30 40 50 60 70

Flux [10−19ergs−1cm2Å−1]

Full (continuum+emission) Model Continuum Model Emission Line Model

6500 7000 7500 8000 8500

Wavelength [Å]

2015 10505 1015 20

Residuals

-4" -2" 0" 2" 4"

-4"

-2"0"2"4"

10 kpc HST ACS F814W

-4" -2" 0" 2" 4"

-4"

-2"0"2"4" HSC gri

5 0 5

Arcseconds

[OII] [NeIII]H8 H² Hδ Hγ Hβ OIII OIII

H10 H9 H8 K H Hδ G Hγ Hβ

ID: 209377 Mask: 5 at z=0.7470

0 50 100 150 200 250

Flux [10−19ergs−1cm2Å−1]

Full (continuum+emission) Model Continuum Model Emission Line Model

6500 7000 7500 8000 8500

Wavelength [Å]

4030 20100 1020 3040

Residuals

Figure 2. Images and spectra of three example quiescent galaxies from the LEGA-C sample, selected to span a range in emission line flux for demonstration of fitting; most galaxies in the sample do not exhibit significant emission. Images from HST ACS COSMOS mosaics and gri color images from the Hyper-Suprime Cam SSP public data release. The position and width of the LEGA-C slit as well as the physical scale are indicated on the HST image. The top panel in each row shows the 2D LEGA-C spectrum, with the location of spectral absorption and emission features, including the measured rotation, indicated with blue and red lines. Emission line features are labeled above the galaxy spectrum and continuum features are indicated below. One-dimensional optimally extracted spectra are included in the middle panel to demonstrate the continuum plus emission-line modeling. Best-fit continuum models are indicated by red lines, emission lines, where detected, are indicated by blue lines, and the combined model by purple lines. Residuals from the 1-D fit are included in the bottom panel. In this work, this procedure is repeated separately on all rows with sufficient S/N in the 2D spectra.

Spectra (MILES) (S´anchez-Bl´azquez et al. 2006) empir- ical stellar spectra combined usingGirardi et al.(2000) isochrones. We extend the rotation curve measurements

in the outer rows (with S/N < 2) by fixing the velocity dispersion to σ = 150 km s⁻¹ and allowing the normal- ization and velocity offsets to vary for both stellar and

-400 -200 0 200 400

5 kpc 117400

Vσ

5 kpc 209377

5 kpc 129631

5 kpc 238580

5 kpc 151161

-400 -200 0 200 400

5 kpc 174876

5 kpc 120372

5 kpc 211767

5 kpc 110509

5 kpc 207735

-400 -200 0 200 400

5 kpc 210564

5 kpc 205742

5 kpc 156055

5 kpc 108227

5 kpc 107489

-400 -200 0 200 400

5 kpc 236682

5 kpc 211736

5 kpc 118552

5 kpc 128311

5 kpc 205655

-400 -200 0 200 400

5 kpc 127986

5 kpc 205715

5 kpc 207825

5 kpc 205548

5 kpc 139772

-400 -200 0 200 400

5 kpc 132048

5 kpc 211157

5 kpc 107468

5 kpc 128133

5 kpc 132394

-1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 -400

-200 0 200 400

5 kpc 97994

-1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 5 kpc 134077

-1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 5 kpc 239469

-1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 5 kpc 126950

-1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 5 kpc 207081

Distance from Center [Arcseconds]

Velocity [km/s]

log(M / M

) > 10.99

Velocity Dispersion [km/s]

Figure 3. Stellar rotation curves (black) and velocity dispersion profiles (red) for the 35 highest mass (log M∗/M > 11) quiescent galaxies, ordered by increasing V5. The rotational velocity is defined as the velocity of the best-fitting arctangent function (indicated by the gray solid lines) at a radius of 5 kpc (indicated by the black bars) from the central pixel.

-400 -200 0 200 400

5 kpc 116829

Vσ

5 kpc 205765

5 kpc 127394

5 kpc 240174

5 kpc 125213

-400 -200 0 200 400

5 kpc 179040

5 kpc 140050

5 kpc 126552

5 kpc 166806

5 kpc 168334

-400 -200 0 200 400

5 kpc 132660

5 kpc 132399

5 kpc 121033

5 kpc 105208

5 kpc 208824

-400 -200 0 200 400

5 kpc 132904

5 kpc 111188

5 kpc 80755

5 kpc 204895

5 kpc 139851

-400 -200 0 200 400

5 kpc 125244

5 kpc 165871

5 kpc 127946

5 kpc 169134

5 kpc 237641

-400 -200 0 200 400

5 kpc 172758

5 kpc 120488

5 kpc 94494

5 kpc 117692

5 kpc 120758

-1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 -400

-200 0 200 400

5 kpc 112534

-1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 5 kpc 118682

-1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 5 kpc 124233

-1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 5 kpc 131013

-1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 5 kpc 134391

Distance from Center [Arcseconds]

Velocity [km/s]

10.73 < log(M / M

) _10.99

Velocity Dispersion [km/s]

Figure 3. (Continued) Stellar rotation curves (black) and velocity dispersion profiles (red) for the 35 intermediate mass (10.7 < log M∗/M≤ 11) quiescent galaxies.

-400 -200 0 200 400

5 kpc 112200

Vσ

5 kpc 119700

5 kpc 237048

5 kpc 90664

5 kpc 151258

-400 -200 0 200 400

5 kpc 209933

5 kpc 205158

5 kpc 173144

5 kpc 178560

5 kpc 183066

-400 -200 0 200 400

5 kpc 133833

5 kpc 126143

5 kpc 108472

5 kpc 208622

5 kpc 204775

-400 -200 0 200 400

5 kpc 123161

5 kpc 129947

5 kpc 139990

5 kpc 126639

5 kpc 205760

-400 -200 0 200 400

5 kpc 208364

5 kpc 87345

5 kpc 146213

5 kpc 206499

5 kpc 117010

-400 -200 0 200 400

5 kpc 88863

5 kpc 167056

5 kpc 139423

5 kpc 124875

5 kpc 139384

-1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 -400

-200 0 200 400

5 kpc 135149

-1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 5 kpc 121482

-1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 5 kpc 182090

-1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 5 kpc 110805

Distance from Center [Arcseconds]

Velocity [km/s]

log(M / M

) _10.73

Velocity Dispersion [km/s]

Figure 3. (Continued) Stellar rotation curves (black) and velocity dispersion profiles (red) for the lowest mass (log M∗/M<

10.7) sample of quiescent galaxies in LEGA-C.

gas templates and find stable results to S/N & 1.2 per pixel. This yields line-of-sight velocity measures out to an average of 8.8 kpc or 2.7R_e. We verify that fixing the velocity dispersion to the nearest measured value does not significantly alter the measured rotational ve- locities, on average leading to a 2% (∆V = 1.2 km s⁻¹) offset, which is well within the measurement uncertainty.

Emission line templates are treated as a single kine- matic component, but the normalization of each line (Balmer lines: H10, H9, H8, H, Hδ, Hγ, Hβ, Hα;

[NeV], [NeVI], [NeIII], and [OII] and [OIII] doublets) is a free parameter in the fit. The optimally extracted 1-D spectra and best-fit models are shown in Figure 2 for three galaxies with increasing emission line compo- nents. These galaxies are representative (e.g. in S/N), but are selected to demonstrate the necessity of includ- ing emission lines in the kinematic fits and the ability of the data to identify emission lines as they fill in broader absorption features. The majority of galaxies in the sample do not have detected emission lines. Images of each galaxy are shown on the left from the COSMOS HST v2.0 ACS Mosaics (top) (Koekemoer et al. 2007) and gri composite color images from the Hyper-Suprime Cam Subaru Strategic Program (HSC-SSPAihara et al.

2017). The two-dimensional spectrum is included in the top panel for each galaxy, with the best-fit rotation curve derived from stellar kinematics at the position of a num- ber of strong absorption (red lines) and emission (blue lines) features overplotted. The middle panel shows the one-dimensional optimally extracted spectrum for each galaxy with the best-fitting continuum model in red and for the second and third galaxies the emission line and total models in blue and purple. The bottom panel in each row shows the residuals from the fit, which are minimal and in most cases uncorrelated.

These fits yield spatially resolved line-of-sight stellar and gas velocity and velocity dispersion profiles along the N-S slits. Although emission lines are present due to residual ionized gas (primarily the [OII] doublet) in a subset of this quiescent galaxy sample, we focus our analysis on the kinematics derived from fitting the stel- lar continuum of each galaxy. Measured stellar rotation curves are shown in Figure 3, in which velocity of the stellar component is indicated by black points and stellar velocity dispersion profiles are shown in red. Rotation curves are plotted in order of increasing rotational ve- locity, separated by page in decreasing mass bins. We fit the rotation curves with an arctangent model and define the line-of-sight rotational velocity (V5) of a galaxy as the value of the best-fitting arctangent at a radius of 5 kpc along the slit. This distance is not corrected for inclination or slit misalignment. We define the central velocity dispersion (σ0) as the velocity dispersion mea- sured in the central pixel (0.205”), which is set as the brightest pixel in the spatial profile. Uncertainties in V5

are estimated by bootstrap resampling within the veloc- ity errors and errors in velocity dispersion are formal un-

certainties estimated by PPXF, with a small correction to underestimated formal errors based on the measured relationship between the measured S/N and formal er- rors.

We adopt this definition of rotational velocity within a fixed physical aperture for two primary reasons. First, the effects of seeing will be similar within a fixed phys- ical radius as opposed to an aperture that scales with the galaxy size. The 5 kpc aperture is used because it is the approximate extent of the shortest LEGA-C rotation curves, and therefore requires minimal extrap- olation. Secondly, utilizing a fixed aperture allows for comparison with galaxies at low redshift in §4 within the same physical region of the galaxy and will be less sensitive to differing apertures due to real size evolution in the galaxy populations. We discuss the impact of this choice of aperture, including the effects of adopting an evolving aperture or utilizing the maximum observed velocity, in the AppendixB.

Given that the effective seeing, including atmospheric and alignment effects, is comparable to the spatial ex- tent of the galaxies themselves (FWHM ∼1.0⁰⁰≈7 kpc) the effects of beam smearing will be significant and kine- matic measurements at each pixel (0.205”) are not in- dependent. This results in shallower than intrinsic ro- tation curves and elevated line-of-sight velocity disper- sions. Dynamical modeling that accounts for aperture and beam smearing effects, which is common in the anal- ysis of emission line kinematics at high redshift (e.g., Vogt et al. 1996,1997;Weiner et al. 2006;Kassin et al.

2007; Simons et al. 2015, 2016; Price et al. 2016; Si- mons et al. 2016;Wuyts et al. 2016;Harrison et al. 2017;

Straatman et al. 2017) can reconstruct the intrinsic ro- tation and velocity dispersion profiles, given modeling assumptions, for direct comparisons with present-day galaxy samples. Such modeling efforts are in prepara- tion (van Houdt et al. in prep), but beyond the scope of the current paper; here we focus on the directly mea- sured rotation (V5) and rotational support (|V5|/σ0). In

§4we reconstruct the rotation and dispersion profiles of local galaxies as they would be observed with LEGA-C at z ∼ 0.8.

3. STELLAR ROTATION IN QUIESCENT LEGA-C GALAXIES

In this section we investigate trends of stellar rotation and rotational support with other properties of massive quiescent galaxies. We specifically focus on stellar mass, with which rotational support has been demonstrated to depend in z ∼ 0 galaxies, and on the two photometric measures that have been used to assess the “disk-like”

nature of massive quiescent galaxies at high redshifts:

projected axis ratio and S´ersic index (e.g., van der Wel et al. 2011; Chevance et al. 2012; Chang et al. 2013;

Cappellari 2016; Graham et al. 2018). We note that the measured |V |/σ will depend on projection effects, which is particularly important in interpreting trends in

Table 1. Properties of Quiescent Galaxies in the LEGA-C Sample

ID zspec log(^M?

M) Re b/a PA n V5 VRe Vmax σ0

[arcsec] [degrees] [km/s] [km/s] [km/s] [km/s]

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11)

80755 0.7325 10.9 1.05±0.02 0.51±0.00 27.54±0.45 3.90±0.08 -68.6±4.6 -61.9±4.4 -116.4±9.7 197.0±8.4 87345 0.6226 10.7 0.67±0.00 0.54±0.00 -35.46±0.26 3.13±0.04 89.7±2.1 51.9±2.2 138.7±2.1 157.5±5.2 88863 0.8124 10.4 0.20±0.00 0.91±0.01 -32.43±0.61 2.62±0.06 -99.0±-99.0 -99.0±-99.0 -99.0±-99.0 96.8±7.1 90664 0.7480 10.2 0.07±0.00 0.50±0.01 -5.14±1.50 4.66±0.17 -13.2±11.8 -3.6±6.2 -12.9±6.3 140.1±17.8 94494 0.7401 10.9 0.49±0.00 0.95±0.00 39.50±0.30 3.72±0.05 92.0±1.0 63.1±0.7 165.3±1.4 215.5±3.9 97994 0.9821 11.2 0.57±0.01 0.64±0.01 19.79±0.48 2.61±0.06 115.8±6.1 83.6±4.4 189.3±8.9 236.6±17.6 105208 0.9345 10.8 0.64±0.02 0.85±0.01 31.85±0.68 5.74±0.20 -31.1±10.5 -27.4±9.3 -55.3±20.7 203.0±22.9 107468 0.9178 11.1 0.21±0.00 0.24±0.00 24.21±0.35 2.27±0.03 -97.8±3.0 -15.1±1.1 -122.4±1.9 234.0±6.2 107489 0.8383 11.1 0.32±0.00 0.44±0.00 26.75±0.26 2.29±0.03 29.8±4.0 7.6±1.0 60.5±2.0 383.6±7.0 108227 0.9603 11.4 1.66±0.03 0.55±0.01 -31.66±0.54 1.30±0.03 -28.0±5.7 -43.5±9.8 -59.2±24.0 263.9±17.0 108472 0.6671 10.6 0.14±0.00 0.56±0.01 8.52±0.68 3.72±0.07 -69.5±2.1 -12.2±0.4 -139.8±0.8 160.8±4.2 110509 0.6671 11.0 0.99±0.01 0.95±0.00 33.81±0.23 3.76±0.04 23.9±1.5 29.6±2.0 41.6±5.8 217.4±5.0 110805 0.7292 10.6 0.47±0.00 0.21±0.00 12.40±0.22 0.55±0.01 -151.5±3.0 -70.4±2.7 -178.8±2.3 172.7±7.5 111188 0.9164 10.9 0.43±0.01 0.58±0.01 11.23±0.65 5.62±0.19 -53.3±9.5 -46.3±5.7 -55.7±13.5 180.1±10.1 112200 0.8279 10.6 0.29±0.01 0.86±0.01 36.97±0.77 4.21±0.15 3.5±8.5 1.3±5.1 3.2±7.1 151.6±13.7 112534 0.9837 11.0 0.34±0.01 0.50±0.01 -21.17±0.60 1.89±0.06 -107.5±7.8 -45.0±5.5 -159.7±7.3 297.3±19.9 116829 0.6683 10.8 0.45±0.00 0.70±0.00 22.15±0.29 2.46±0.03 -1.5±2.0 -1.5±1.7 -1.5±2.0 162.3±6.7 117010 0.6766 10.4 0.40±0.01 0.54±0.01 -27.19±0.61 4.45±0.13 99.0±3.4 36.3±1.8 135.3±2.7 158.1±5.3 117400 0.6687 11.3 0.80±0.01 0.79±0.00 40.78±0.28 4.85±0.06 1.3±3.7 1.2±3.5 3.1±5.3 258.2±4.7 117692 0.6753 10.8 0.63±0.01 0.48±0.00 -22.70±0.34 4.13±0.07 -100.9±2.4 -54.2±1.4 -189.4±2.6 185.4±7.3

Note—Table 1 will be published in its entirety in the machine-readable format. A portion is shown here for guidance regarding its form and content.

This table includes measured properties of the galaxies included in this sample from the year one LEGA-C dataset. All galaxies included in this table are well-aligned with the N-S VIMOS slits (|PA| < 45^o), quiescent based onMuzzin et al.(2013b) U-V and V-J rest-frame color cuts, have reliable morphological parameters measured from ACS F814 images, and represent single Virialized systems. Columns: (1) ID from theMuzzin et al.(2013a) UltraVISTA DR1 v4.1 catalogs; (2) spectroscopic redshift; (3) log Stellar Mass assumingChabrier(2003) IMF; (4) S´ersic semi-major axis; (5) projected axis ratio; (6) major axis position angle; (7) S´ersic index; (8) average line-of sight rotational velocity measured at 5kpc; (9) average line-of sight rotational velocity measured at the effective radius; (10) average line-of sight rotational velocity measured at the maximum extent; (11) velocity dispersion measured in the central pixel in the spatial dimension.

projected axis ratio. Therefore in addition to |V₅|/σ0, we introduce (V5/σ0)^∗following e.g. Binney(1978);Davies et al.(1983), which is defined as the V /σ normalized by the (V /σ)O for an oblate, isotropic model and should be largely independent of projection effects. We adopt the approximation V /σ ≈ p/(1 − ) from Kormendy (1982). Following this definition,

(V5/σ0)^∗= (|V5|/σ0)

p/(1 − ). (1)

Figure 4 shows rotational velocity (V₅) of galaxies in the top row, velocity dispersion in the second row, rotational support (|V5|/σ0) in the third row, and in the bottom row rotational support with a correction for projection effects, (V₅/σ₀)^∗, as a function of stellar mass (left), projected axis ratio (center), and S´ersic in- dex (right). Average uncertainties on the measurements are indicated by errorbars in the upper right corners of each panel. Running median and mean are indicated by red dashed and blue solid lines respectively for bins

with greater than three data points. Errors on the mean are estimated in each bin via jackknife resampling. In each case these trends are best described as scatter be- tween no rotational support and a maximum value that depends on the property plotted on the horizontal axis.

This leads to measured (anti-)correlations, for which we quote the Pearson correlation coefficient in the upper left corner of each panel.

In the left panels, we see that more massive galaxies exhibit lower rotational support (|V5|/σ0 or (V5/σ0)^∗) than less massive galaxies. This is also evident in the local universe (e.g., Emsellem et al. 2011). We will re- turn to this trend in Figure5, where we also include in- formation about galaxy morphology in the same panel.

We emphasize that this is primarily due to the known correlation between stellar mass and velocity dispersion, the mass Faber & Jackson (1976) relation (left panel, second row); rotational velocities alone do not exhibit a strong correlation with stellar mass (top left panel).

However, at all masses there is at least a small fraction of galaxies that are observed to have very little rota-

tional support. Some of this is an observational effect:

beam smearing, inclination, and slit misalignment di- minish ordered rotation and increase observed velocity dispersions and we expect this to preferentially impact smaller galaxies. We investigate these effects in greater detail in Section4.2.

Another key result of our measurements is that galax- ies that are flat in projection generally show rotation in their stellar body, whereas round galaxies do not (top center panel in Figure 4). This is well-understood as largely due to a combination of intrinsic elongation and projection effects (e.g.,Cappellari et al. 2007;Emsellem et al. 2007,2011;Fogarty et al. 2014,2015;van de Sande et al. 2017). This trend is tightened when rotational support is compared to dispersion support in the cen- tral pixel ( third row, center panel in Figure 4), with a Pearson correlation coefficient r = −0.41. This is primarily a trend in rotational velocity, not velocity dis- persion (see middle panel, second row). There is a sub- set of elongated galaxies that show little rotation ( 3 of 25 galaxies with b/a < 0.5 have |V5|/σ0 < 0.1). The nature of these galaxies remains to be determined, but perhaps they are not unlike NGC4550, which does not show net rotation but has been demonstrated to con- sist of two counter-rotating disks (Johnston et al. 2013).

This overall trend implies that the distribution of pro- jected axis ratios for a population of quiescent galax- ies will be a decent estimate of the overall observed degree of rotational support. However, for any individ- ual galaxy with an observed axis ratio of b/a & 0.6 a significant fraction of galaxies will still have significant rotation and spatially resolved kinematics will be nec- essary to distinguish between pressure and rotationally supported systems.

Intriguingly, although both velocity (top right panel) and rotational support ( third row, right panel) exhibit a statistically significant correlation with S´ersic index, the mean relation turns over exactly at the S´ersic in- dex where one would expect the anti-correlation to be strongest. Although the numbers are small, the mean rotational velocity of galaxies that would be classified as disk-like based on their concentrations (n < 2.5) is not elevated (h|V5|/σ0i = 0.33, median= 0.34) com- pared to the overall average (h|V5|/σ0i = 0.31). This trend is strongest for the highest mass quiescent galax- ies (log M_?/M > 11), for which the n < 2.5 average h|V5|/σ0i = 0.20 versus overall h|V5|/σ0i = 0.25. These massive galaxies are the most extended, and therefore the least affected by beam smearing, and yet this trend is contrary to expectations. Larger samples, such as the full 4 year LEGA-C sample, will likely include a larger number of n < 2.5 galaxies and allow for a more statisti- cally significant assessment of these trends. Regardless, we emphasize that measuring the S´ersic index of an in- dividual quiescent galaxy cannot determine whether it is rotationally supported. Overall, S´ersic index is anti-

correlated with rotational support, with a weaker Pear- son coefficient r = −0.41).

Although measured |V |/σ will likely be sensitive to projection effects, (V5/σ0)^∗, which normalizes out ex- pected V /σ based on projected axis ratios for an oblate, isotropic model, should be largely independent of projec- tion effects. The bottom row of Figure4shows (V5/σ0)^∗ as a function of stellar mass, projected axis ratio, and S´ersic index in the bottom panels. Although all quan- tities are still correlated with this measure of rotational support, it is clear that a significant fraction of the cor- relation with projected axis ratio was covariance of the variables; once the projection effects are removed pro- jected axis ratio exhibits a mild correlation with rota- tional support (r = 0.25). This remaining correlation is likely driven by the four round (b/a > 0.8) galax- ies with high (V5/σ0)^∗ that are not well approximated by isotropic oblate rotators. We note that although inclination and projection effects can account for some of the anti-correlation between |V₅|/σ0and S´ersic index, the weak anti-correlation remains between (V5/σ0)^∗and S´ersic n. We reiterate that this sample includes very few low S´ersic index galaxies and although we caution again the use of S´ersic index to characterize individual galax- ies, we do not have the statistics to characterize this trend at low S´ersic indices.

Figure 5 shows rotational support (|V5|/σ0) versus stellar mass, but now with symbols that reflect mor- phologies. Symbol sizes correspond to logarithmically scaled galaxy effective radii, symbol axis ratios and po- sition angles reflect the projected galaxy shapes and ori- entations. Symbol colors correspond to S´ersic index.

The average uncertainty is indicated by the errorbars in the upper right and the mean trend, as calculated in Figure 4, is indicated by the gray band. Here we can clearly identify massive galaxies with seemingly incon- sistent morphologies and measured kinematics: galaxies with little observed rotational support, but low S´ersic indices (purple colors) as well as others with high S´ersic indices (orange colors) and high |V₅|/σ₀.

Our measured |V₅|/σ0 is likely to be an underesti- mate due to a number of observational effects such as rotational velocities contributing to central velocity dis- persions and decreasing measured line-of-sight velocities due to inclination. Therefore, galaxies with low mea- sured rotational support may in fact be revealed to be intrinsically fast-rotators with full modeling; however galaxies that are observed to be rotating quickly cannot be slow-rotators. Given this observational ambiguity we refrain from using the terms “fast” and “slow” rotators, but return to quantifying the observational biases in the following section.

Our kinematic measure |V5|/σ0is not directly compa- rable to the classifiers used for present-day galaxies as seeing, slit-misalignment, and other observational effects are not taken into account. However, the trends in Fig- ure4are very similar to those observed for present-day

0 50 100 150 200

Velocity at 5 kpc (|

V

r = -0.11

Median

Mean r = -0.49 r = -0.39

50 100 150 200 250 300 350 400 450

Velocity Dispersion [km/s]

r = 0.71 r = -0.09 r = -0.01

0.0 0.2 0.4 0.6 0.8 1.0

|

V

/σ

r = -0.30

0.2 0.4 0.6 0.8 1.0 r = -0.41

0 1 2 3 4 5 6

r = -0.41

10.0 10.5 11.0 11.5

log Stellar Mass [M_¯] 0.0

0.5 1.0 1.5 2.0

(

V

/σ

r = -0.18

0.2 0.4 0.6 0.8 1.0 Axis Ratio

r = 0.25

0 1 2 3 4 5 6

Sersic Index r = -0.14

Figure 4. Rotational velocity (|V5|, top row), central velocity dispersion (σ0, second row), rotational support (|V5|/σ0, third row), and rotational support normalized by the expectation for an oblate rotator given the measured projected axis ratio ((V5/σ0)∗, bottom row) in quiescent LEGA-C galaxies versus stellar mass (left), projected axis ratio (middle), and S´ersic index (right). Individual galaxies are indicated by small gray symbols, median and mean trends are indicated by red dashed and blue solid lines and symbols, respectively. The strongest correlation exists between stellar mass and velocity dispersion, or the

“mass” Faber-Jackson relation. Projected axis ratio exhibits the strongest anti-correlation with |V5|/σ0and unlike S´ersic index, the population average with |V5|/σ0 does not flatten out at elongated axis ratios in this sample. When projection effects are minimized with (V5/σ0)^∗, this removes significant correlations with projected axis ratios, suggesting roughly similar correlations between rotational support and stellar mass, axis ratio, and S´ersic index.

10.4 10.6 10.8 11.0 11.2 11.4 11.6 log Stellar Mass [M

]

0.0 0.2 0.4 0.6 0.8 1.0 1.2

|

V

/ σ

0 1 2 3 4 5 6

Sersic Index

Figure 5. Rotational support (|V5|/σ0) versus stellar mass for the LEGA-C sample of massive, quiescent galaxies. Sym- bol size indicates the galaxy effective radii (in log scale) and position angles and axis ratios of symbol ellipses reflect those of the galaxies. Symbol color indicates S´ersic index. The mean relation is indicated by gray band and average uncer- tainty is indicated by errorbars in the upper right corner.

The majority of high-mass galaxies have minimal rotational support, even when their S´ersic indices are disk-like, how- ever there are several high-mass galaxies with significant ro- tation. Below log M/M. 11.2 galaxies exhibit a range in rotational support and smaller and more elongated galaxies consistently show higher measured |V5|/σ0.

galaxies (e.g., Emsellem et al. 2011), and we conclude that at all cosmic times since at least z ∼ 1 the quiescent galaxy population consists of galaxies with low and high degrees of rotational support that reflect their intrinsic structure (spheroidal/triaxial and disk-like/oblate, re- spectively). At the same time, among the 10 most mas- sive galaxies with stellar masses > 2 × 10¹¹M, only 2 show evidence for rotation. This is suggestive that the only way that galaxies can grow to such large masses is by a mechanism that reduces the angular momentum, that is, dissipationless merging. In the following section, we analyze the CALIFA dataset to further explore the question of quantifying this evolution.

Although we focus on the quiescent sample only for this paper, we note that as expected, the star-forming and quiescent galaxy populations differ in dynamics as well as stellar populations. Figure6shows the observed rotational support (|V5|/σ0) versus stellar mass for all galaxies with photometric axes within 45^o of the N- S slits. Quiescent galaxies are indicated by red cir- cles and star-forming galaxies by blue diamonds. The star-forming galaxies have more rotational velocity than

10.5 11.0 11.5

log Stellar Mass [M

] 0.0 0.2

0.4 0.6 0.8 1.0 1.2 1.4 1.6

|

V

/σ

Quiescent Star-forming

0 10 20 N

Figure 6. Observed rotational support of LEGA-C galaxies versus stellar mass for star-forming (blue diamonds) and qui- escent galaxies (red circles). Average uncertainties, shown as blue and red errorbars in the upper right, are higher for the star-forming galaxies (∼ 0.1) than for quiescent galax- ies (∼ 0.04) in the LEGA-C sample. The right panel in- dicates the histograms in rotational support between the star-forming and quiescent populations; the distributions are overlapping but on average star-forming galaxies show higher V5/σ0 than quiescent galaxies overall and at fixed mass.

quiescent galaxies, as found in the local Universe (e.g.

Cortese et al. 2016). Figure 1demonstrates the known bimodality of these two populations in size and specific star formation rate, this figure provides the first evi- dence for dynamical bimodality at high redshift based on stellar kinematics. The two populations overlap in observed phase space, however their distributions dif- fer significantly (see histograms in the right panel). A two-sample K-S test rejects the possibility that they are drawn from the same distribution with a p = 1 × 10⁻¹⁰, or p = 4 × 10⁻⁸ for massive log M_?/M > 10.4 galax- ies. Average values of errors on V₅/σ₀ for the star- forming and quiescent sub-samples are indicated by blue and red errorbars in the upper right corner. Uncertain- ties in the |V5|/σ0values, especially for the star-forming population, contribute significantly to the broadening of the distribution. Therefore, this discrepancy may be stronger in the intrinsic properties of the two popula- tions. We leave the analysis of the dynamics of star- forming galaxies and of the joint population to future studies (Straatman et al. in prep, van Houdt et al. in prep).

4. CALIFA STELLAR KINEMATICS AND THE REDSHIFT EVOLUTION OF ROTATIONAL

SUPPORT

The Calar Alto Legacy Integral Field Area (CALIFA) survey provides an excellent census of the spectroscopic

E0 E1 E2 E3 E4 E5 E6 E7 S0 S0a Sa Sab Sb Morphological Type

0 2 4 6 8 10 12 14 16

Number of CALIFA Quiescent Galaxies

Figure 7. Morphological distribution of the 91 CALIFA galaxies determined to be quiescent based on EW(Hα) < 3˚A in the stellar kinematics sample.

properties of local (0.005 < z < 0.03) galaxies of all morphological and spectral types (S´anchez et al. 2012;

Walcher et al. 2014). The CALIFA team has promptly provided reduced data products in public data releases in addition to derived spectroscopic properties. For this project we include CALIFA galaxies from Data Release 3 (DR3,S´anchez et al. 2016), stellar kinematics maps from Falc´on-Barroso et al. (2017), and spectroscopic classifi- cations based on ionized gas lines fromCano-D´ıaz et al.

(2016). Using this dataset, we use intensity, stellar ve- locity, and stellar velocity dispersion fields in two spatial dimensions and extract profiles along a variety of axes and replicate the LEGA-C kinematic analysis on a local sample, quantifying intrinsic properties and simulating the effects of seeing on the measured LEGA-C rotation curves.

4.1. The CALIFA Dataset

Of the 667 galaxies in the full DR3, 300 are included in the Falc´on-Barroso et al. (2017) analysis of stellar kinematics. This sample of galaxies, which have been observed with both low (V500) and medium (V1200) resolution gratings, is deemed to be representative of the full CALIFA sample in magnitude, size, and redshift and spans a wide range of morphological types. As in the LEGA-C sample, Falc´on-Barroso et al. (2017) remove strongly interacting galaxies from this kinematic sam- ple. Falc´on-Barroso et al.(2017) analyze IFU datacubes for each galaxy, which are Voronoi binned to S/N ∼ 20 and the stellar kinematics are measured in each bin by fitting a combination of stellar templates convolved with a gaussian line-of-sight velocity dispersion. These

fits yield maps of velocity and velocity dispersion at each spaxel, which the authors provide on the CAL- IFA website (http://califa.caha.es/?q=content/

science-dataproducts). Falc´on-Barroso et al. (2017) also provide stellar masses assuming a Chabrier(2003) IMF and effective radius, ellipiticity, and position angle determined from the outer parts of the galaxies in SDSS imaging as described inWalcher et al.(2014).

We further limit our analysis to quiescent galaxies fol- lowing the classifications ofCano-D´ıaz et al.(2016), who determine Hα-based star formation rates and use ion- ized gas lines to differentiate amongst dominant ioniza- tion sources using EW(Hα) and the Kewley demarca- tion limit (Kewley et al. 2001) in the Baldwin-Philips- Terlevich (BPT) diagram (Baldwin et al. 1981). Cano- D´ıaz et al. (2016) identify each CALIFA galaxy as ei- ther “Star-forming”, “AGN”,“Retired”, or in ambiguous cases “Undefined”. The Cano-D´ıaz et al.(2016) study included a representative sample of 535 galaxies that had been observed by February, 2015; and therefore does not completely overlap with theFalc´on-Barroso et al.(2017) sample. We classify the remaining ten galaxies by eye using the two-dimensional star-formation maps provided in the CALIFA DR3. For the most conservative com- parison with the current study, we limit our analysis to the quiescent or “Retired” (EW(Hα) < 3˚A) sample of galaxies, based on their spectroscopic properties. Only 4 galaxies are classified as Retired by eye and we verify that excluding these galaxies does not significantly im- pact any of the conclusions in this paper. For maximum consistency in stellar population modeling, we compare stellar masses with those derived byBrinchmann et al.

(2004) for the subset of these galaxies which also fall in the spectroscopic SDSS DR7 sample. These fits are also based on aperture photometry and are analyzed using similar methodology to our modeling of the UltraVISTA photometry. We find that CALIFA stellar masses are higher than those derived by Brinchmann et al.(2004) by a median of 0.16 dex for the retired galaxy popu- lation. We perform a linear regression to this subset and apply this correction to the CALIFA-derived stellar masses. The final sample includes 91 galaxies across a range of morphological types, from E0 to Sb as shown in Figure7.

4.2. Simulating LEGA-C observations with CALIFA datacubes

For each galaxy in the quiescent CALIFA sample, we extract the intrinsic intensity I1D(x), velocity V (x), and velocity dispersion σ(x) profiles along lines pass- ing through the maximum of the intensity map of the galaxy. These 1D profiles are measured along the pub- lished galaxy photometric position angles, as determined by Walcher et al. (2014) from galaxy outskirts in the SDSS imaging. Rotation curves are fit with arctangent functions and rotational velocity at 5 kpc and central velocity dispersion are measured as for the LEGA-C

0.0 0.5 1.0 1.5

|

V

/σ

10

10

10

10

10

( V/ σ

) / ( V/ σ )

Slit Misalignment

Median = 0.92

0.0 0.5 1.0 1.5

|

V

/σ

Beam Smearing

Median = 0.46

0.0 0.5 1.0 1.5

|

V

/σ

Both

Median = 0.38

Figure 8. Ratio of simulated “observed” to intrinsic rotational support (|V5|/σ0) versus the intrinsic value due to slit misalignment (left panel), beam smearing (center panel), and the combined effects (right panel). Galaxies with minimal rotational support (|V5|/σ0 < 0.1) are indicated by small symbols and those with higher |V5|/σ0 by large symbols. Beam smearing is the dominant effect, decreasing the observed |V5|/σ0by an median factor of ∼2.2, while slit misalignment decreases the value by ∼8%.

10.5 11.0 11.5

log Stellar Mass [

M

10^-1 10⁰ 10¹ 10²

( V/ σ

)

( V/ σ )

|

V

/σ

< 0 . 1

|

V

/σ

0 . 1

1.0 10.0

Effective Radius [kpc]

Figure 9. Trends in the ratio of observed to intrinsic |V5|/σ0 from the CALIFA simulations with stellar mass in the left panel and effective radius in the right panel. Small symbols indicate galaxies with |V5|/σ0 < 0.1, which are most sensitive to this relative metric. The running mean relations are indicated by blue dashed (all galaxies) and red solid (|V5|/σ0 > 0.1) lines.

As expected, the blurred rotational support preferentially impacts the least massive and most compact galaxies because of the relative size of the PSF and the galaxy extent.

dataset. These values correspond to the intrinsic V5and σ0values.

We use the two-dimensional intensity and kinematic maps, spatially subsampled by a factor of 100, to sim- ulate the observational effects of the misaligned 1” slits (∼ 7.5kpc), 0.205” pixels (∼ 1.5kpc), and seeing charac- teristic of the LEGA-C observations. Slit misalignment in the LEGA-C survey, which in this study is limited to within 45^o of the North-South slits, is simulated by extracting one-dimensional profiles along the closer of

the horizontal or vertical directions. The intensity in two-dimensional position-velocity space can be defined as:

I_3D(x, y, v) = I_2D(x, y) exp

"

−(v − V (x, y))² 2σ(x, y)²

# . (2) The effects of seeing are then simulated by convolv- ing this intensity (I3D) field with a two-dimensional Moffat profile. For this we adopt a uniform value of